Python script to take GEDI level 2 data and convert variables to a geospatial vector format. Usage `python gedi_to_vector.py <path> --variables [<var1>,<var2>,...,<varN>] --outFormat <extension> --filterBounds [<W>,<S>,<E>,<N>] --verbose`

gedi_to_vector.py

import os
import fire
import h5py
import glob
import tqdm
import numpy as np
import pandas as pd
import geopandas as gpd
# requires fire, h5py, tqdm, numpy, pandas, and geopandas

def gedi_to_vector(file,variables=None,outFormat='CSV',filterBounds=None):
    # open hdf5 file
    data = h5py.File(file,'r')

    # get full file name and extension
    name,_ = os.path.splitext(file)

    # create empty dataframe to append data to
    df = pd.DataFrame()

    # loop over all of the hdf5 groups
    for k in list(data.keys()):
        # if BEAM in the group name
        if 'BEAM' in k:
            # get the geolocation subgroup
            geo = data[k]['geolocation']
            d = {}
            # loop through all of the variables defined earlier
            for var in variables:
                # assign variable array to dict key
                d[var] = np.array(geo[var])
            # convert dict of varaibles to dataframe
            tdf = pd.DataFrame(d)
            # concat to larger dataframe
            df = pd.concat([df,tdf],axis=0,sort=False)

    # check if the the filterBounds is provided
    if filterBounds is not None:
        w,s,e,n = filterBounds # expand list to individual variables
        # select features on X axis
        horizontalMask = (df.longitude_bin0 >= w) & (df.longitude_bin0 <= e)
        # select features on Y axis
        verticalMask = (df.latitude_bin0 >= s) & (df.latitude_bin0 <= n)
        # combines masks to select features that intersect
        spatialMask = verticalMask & horizontalMask
        # grab only masked values within the bounds provided
        df = df.loc[spatialMask]

    # check to make sure that the dataframe has values.
    # if not, then return from function without saving df
    if df.size == 0:
        return

    if outFormat in ['CSV','csv']:
        # save dataframe of parsed variables to CSV file
        df.to_csv('{}.{}'.format(name,outFormat.lower()),index=False)

    else:
        # check if df has the geoinformation
        if ('latitude_bin0' not in df.columns) or ('longitude_bin0' not in df.columns):
            raise KeyError("Geospatial variables 'latitude_bin0' and/or 'longitude_bin0' were not found, "
                "please specify these variables to be extracted when writing to geospatial format")

        # convert to geodataframe
        gdf = gpd.GeoDataFrame(
            df, geometry=gpd.points_from_xy(df.longitude_bin0, df.latitude_bin0))
        # save the geodataframe of variables to file
        gdf.to_file('{}.{}'.format(name,outFormat.lower()))

    return

def main(path,variables=None,verbose=False,outFormat='CSV',filterBounds=None):
    # check if the variables to extract have been defined
    if variables is None:
        raise ValueError("Please provide variables from the GEDI file to convert")

    # if variables have been defined, check if provided in correct datetype
    if type(variables) is not list:
        raise TypeError("Provided variables is not list, please provide argument as '[<var1>,<var2>,<var3>]'")

    # check if filterBounds have been provided and in correct datatype
    if (filterBounds is not None) and (type(filterBounds) is not list):
        raise TypeError("Provided filterBounds is not list, please provide argument as '[W,S,E,N]'")

    # check if the output format provided is supported by script
    availableFormats = ['CSV','SHP','GeoJSON','GPKG','csv','shp','geojson','gpkg']
    if outFormat not in availableFormats:
        raise NotImplementedError('Selected output format is not support please select one of the following: "CSV","SHP","GeoJSON","GPKG"')

    # check if path provided is a file or folder
    if os.path.isfile(path):
        flist = [path]
    else:
        # only search for h5 files in the path provided
        flist = glob.glob(os.path.join(path,'*.h5'))

    if verbose:
        print('\n')
        t = tqdm.tqdm(total=len(flist))

    # loop through the files and do the conversion
    for i,f in enumerate(flist):
        if verbose:
            _, desc = os.path.split(f)
            t.set_description(desc="Processing {}".format(desc))

        gedi_to_vector(f,variables,outFormat,filterBounds)

        if verbose:
            t.update(i+1)

    return

if __name__ == "__main__":
    fire.Fire(main)

Greetings, I am trying to use gedi_to_vector.py to extract height, but not sure what variables from h5 metadata I should use for --variables [...]. Tried [height_bin0], but got error "AttributeError: 'DataFrame' object has no attribute 'longitude_bin0'". Wonder if you could help? Thanks

Greetings, I am trying to use gedi_to_vector.py to extract height, but not sure what variables from h5 metadata I should use for --variables [...]. Tried [height_bin0], but got error "AttributeError: 'DataFrame' object has no attribute 'longitude_bin0'". Wonder if you could help? Thanks

I also had the same problem the first time I used this code. The longtitude_bin0 and latitude_bin should be included intovariable.
variables=['height_bin0', 'longtitude_bin0', 'latitude_bin'].

Wonderful! Thanks for your quick response. It works.

Any thoughts on how to extract the 2-dimensional variables cover_z, pai_z and pavd_z for z = 0 - 29?

@ovanlier that would require a change to the code where at lines 26-31 it is only searching through the geolocation subgroup but would need to expand the search (or allow a search) to all subgroups within the file.

I will try to update the script soon (hopefully today) where you would not have to explicitly call for the latitude_bin0 and longitude_bin0 variables and extract any variable.

@KMarkert thank you for your reply.
The fork submitted by rbavery (found here) allows for extracting variables outside the geolocation subgroup.
Most variables are 1-dimensional, however, variables like cover_z, pai_z and pavd_z are 2-dimensional. The issue lies in attempting to extract the 2D array variables. For example, the 2D variable cover_z has a 1D array for every z for z = 0, 1, 2, ..., 29.
It would be nice to extract cover_z for all z (i.e., cover_0, cover_1, cover_2, ..., cover_29).

Oh, thanks for letting me know! Probably saved me a bunch of time figuring that the data is 2D...the 2D format should be fun to extract but doable. I will have a look and update when I have something.

Great job K.! keep up the good work.

Impressive R package for all your GEDI processing: https://github.com/carlos-alberto-silva/rGEDI

@ovanlier, agreed. There are some nice packages coming out to interface with the GEDI data. Here is a Python package too: https://github.com/EduinHSERNA/pyGEDI

I looked into the Python package and it seems that there is the ability to extract out data, plot, and do whatever else is needed. So, it looks like this script may be obsolete....I would suggest using the packages for analysis since they are maintained and have more features.

Hi
I m getting some kind of syntax error as shown in attached preview . Please help me in this regard as it unable to sort out this error from myside